In-order sliding-window aggregation in worst-case constant time

Sliding-window aggregation is a widely-used approach for extracting insights from the most recent portion of data stream. While aggregations interest can usually be expressed as binary operators that are associative, they not necessarily commutative nor invertible. Non-invertible operators, however, difficult to support efficiently. DABA first algorithm sliding-window with worst-case constant time. Prior DABA, best published algorithms would require $$O(\log n)$$ steps per window operation size n—and while strictly in-order streams, this bound could improved O(1) in amortized sense, it was known how achieve an worst case, which critical latency-sensitive applications. In article, besides describing more detail, we introduce new variant, Lite, achieves same time bounds less memory. Whereas requires space storing 2n partial aggregates, Lite only $$n+2$$ aggregates. Our experiments on synthetic and real theoretical findings.